​In non-destructive testing (NDT) 3D images are usually constructed by post-processing data from ultrasonic probes. And, with processing times from a few minutes up to an hour (depending on the number of voxels in the final 3D image), "real-time" imaging is simply not possible. Researchers at List, a CEA Tech institute, sped image processing up by a factor of 30, effectively eliminating a major obstacle to real-time 3D imaging.

The research, which was for a PhD dissertation, leveraged existing technologies used in medical imaging. They tweaked a space-time Fourier transform algorithm to factor in the unique characteristics of non-destructive testing, such as a variety of waves, the presence of interfaces between the probe and part to inspect, and the reflection of waves off of the interfaces. An initial proof-of-concept confirmed that the algorithm was valid by performing the calculations in parallel on multicore graphics processors. The 3D images obtained presented the same level of quality as those produced using conventional methods—and they took just seconds to process.

The algorithm makes it possible to see the geometry of a defect "live" and in 3D from a single sensor position. Another PhD research project is focusing on speeding up data transfer to the CPUs with the goal of building a prototype real-time 3D imager.